vulnhuntr - Zero shot vulnerability discovery using LLMs

项目说明

<di="center">

<img width="250" src=" alt=" Logo">

A tool to identify remotely exploitable using LLMs and static code analysis.

World's first autonomous AI-discovered 0day

</div>

Description

le the power of LLMs to automatically create and analyze entire code call chains starting from remote user input and ending at ser for detection of complex, multi-step, security-bypassing that go far beyond what traditional static code analysis tools are capable of performing. See all the details including the output for all the 0-days here: Protect AI Blog

Found

> [!TIP]
> Found a using ? Submit a report to huntr.com to get $$ and submit a PR to add it to the list below!

> [!NOTE]
> This table is just a sample of the found so far. We will unredact as responsible disclosure periods end.

| Repository | Stars | |
| - | - | - |
| gpt_academic | 67k | LFI, XSS |
| ComfyUI | 66k | XSS |
| Langflow | 46k | RCE, IDOR |
| FastChat | 37k | SSRF |
| Ragflow | 31k | RCE |
| LLaVA | 21k | SSRF |
| gpt-researcher | 17k | AFO |
| Letta | 14k | AFO |

Limitations

Only Python codebases are supported.

Can only identify the following classes:

- Local file include (LFI)
- Arbitrary file o (AFO)
- Remote code execution (RCE)
- Cross site scripting (XSS)
- SQL Injection (SQLI)
- Server side request forgery (SSRF)
- Insecure Direct Object Reference (IDOR)

Installation

> [!IMPORTANT]
> strictly requires Python 3.10 because of a number of bugs in Jedi which it uses to parse Python code. It will not work reliably if installed with any other of Python.

We recommend using pipx or Docker to easily install and run .

Using Docker:

Using pipx:

Alternatively you can install directly from source using poetry:

Usage

This tool is designed to analyze a GitHub repository for potential remotely exploitable . The tool requires an API key and the local path to a GitHub repository. You may also optionally specify a custom endpoint for the LLM service.

> [!CAUTION]
> Always set spending limits or closely monitor costs with the LLM provider you use. This tool has the potential to rack up hefty bills as it tries to fit as much code in the LLMs context window as possible.

> [!TIP]
> We recommend using Claude for the LLM. Through testing we have had better results with it over GPT.

Command Line Interface

Examples

From a pipx install, analyze the entire repository using Claude:

> [!TIP]
> We recommend giving specific files that handle remote user input and scan them indi.

From a pipx install, analyze the /path/to/target/repo/server.py file using GPT-4o. Can also specify a subdirectory instead of a file:

From a docker installation, run using Claude and a custom endpoint to analyze /local/path/to/target/repo/repo-subfolder/target-file.py:

*Experimental*

Ollama is included as an option, however we haven't had success with the open source models structuring their output correctly.

Logic Flow

! logic

LLM summarizes the README and includes this in the system prompt

LLM does initial analysis on an entire file and reports any potential

then gives the LLM a prompt for secondary analysis

Each time the LLM analyzes the code, it requests additional context functions/classes/ from other files in the project

It continues doing this until the entire call chain from user input to ser is complete then gives a final analysis

The final analysis consists of it's reasoning, a proof of concept exploit, and a confidence score

>[!TIP]
> Generally confidence scores < 7 mean there's unlikely a . Confidence scores of 7 mean it should be in, and confidence scores of 8+ mean it is very likely to be a valid .

Output

The tool generates a detailed report of the found in the analyzed files. The report includes:

Initial assessment results for each file.

Secondary assessment results with context functions and class references.

Confidence scores for found.

Logs of the analysis process.

PoC exploit

Below is an example of a report describing a 0-day remote code execution in Ragflow (now fixed):

Logging

The tool logs the analysis process and results in a file named .log. This file contains detailed information about each step of the analysis, including the initial and secondary assessments.

Authors

Dan McInerney: ,

Marcello Salvati: ,

项目信息

开发语言：Python

Star 数：2794

技术要点

该项目属于漏洞知识库方向的开源资源，代码结构清晰，适合安全研究人员学习参考。项目围绕Zero shot vulnerability discovery using LLMs展开，提供了可运行的实现，对理解相关安全机制、工具原理与攻防技术具有较高的研究价值。

从技术实现角度看，此类项目的核心在于对底层原理的深入理解与工程化落地。开发者在研究过程中可以结合官方文档与社区讨论，梳理项目的初始化流程、核心模块职责与数据流转，逐步掌握其中的技术细节与设计思路。

使用与学习建议

1. 下载源码压缩包，解压到本地环境。

2. 根据项目类型准备对应的运行环境与依赖。

3. 按照 README 中的步骤完成配置与部署。

4. 在隔离的测试环境中运行验证，避免影响生产系统。

免责声明

本资源来源于开源社区，仅供安全研究、技术学习与合法授权范围内的攻防演练使用，请勿用于任何非法或侵害他人权益的用途。使用本项目产生的一切后果由使用者自行承担。